如何在 t-sql (Azure Synapse) 的 CAST/CONVERT 中使用字符串函数

Question

我在 Azure Synapse 中有一张表，如下所示，

想要将 Enrolled_period 列查询为 2 个不同的列作为 DATE 数据类型。

我正在尝试以下查询

  select convert(varchar, SUBSTRING(enrolled_period, 2, 12), 23) as startdate
from dbo.test_period;

也尝试拆分数据

select 
REPLACE(PARSENAME(REPLACE(enrolled_period, ', ', '.'), 2), '(', '') AS startdate, 
REPLACE(PARSENAME(REPLACE(enrolled_period, ', ', '.'), 1), ')', '') AS enddate 
from dbo.test_period

但出现错误

从字符串转换日期和/或时间时转换失败。

我该如何解决这个问题？

Answer 1

我创建了一个测试表并尝试了这个，它的工作原理：

请检查：

select 
cast ( SUBSTRING(p,3,10) as date) , cast ( SUBSTRING(p,16,10) as date),p
from test2

Answer 2

在尝试转换/转换为日期之前，最好先测试生成的子字符串是什么样子的。

例如：

declare @enrolled_period varchar(30);
set @enrolled_period = '(''2021-10-11'', ''2021-10-31'')';

select @enrolled_period as enrolled_period
, substring(@enrolled_period, 3, 10) as start1
, substring(@enrolled_period, 17, 10) as end1
, RIGHT(PARSENAME(REPLACE(@enrolled_period, ''', ''', '.'), 2), 10) as start2
, LEFT(PARSENAME(REPLACE(@enrolled_period,  ''', ''', '.'), 1), 10) as end2

select @enrolled_period as enrolled_period
, CAST(SUBSTRING(@enrolled_period, 3, 10) AS DATE) as startdate
, CAST(SUBSTRING(@enrolled_period, 17, 10) AS DATE) as enddate

子字符串的第三个参数是长度，而不是位置。

额外

只是为了展示使用 STRING_SPLIT 或 JSON 来实现这一点是可能的，但会有点矫枉过正。

declare @T table (
  id int identity(1,1) primary key,
  enrolled_period varchar(30)
);

insert into @T (enrolled_period) values
('(''2021-10-11'', ''2021-10-31'')')
;

SELECT t.*, ep.startDate, ep.endDate
FROM @T t
OUTER APPLY (
   select 
    min(try_cast(substring(value,3,10) as date)) as startDate, 
    max(try_cast(substring(value,3,10) as date)) as endDate
   from string_split(t.enrolled_period,',') s
) ep;


SELECT t.*, ep.*
FROM @T t
OUTER APPLY (
  select v.js as period
  , try_cast(json_value(v.js,'$[0]') as date) as startDate
  , try_cast(json_value(v.js,'$[1]') as date) as endDate
  from (
    select replace(replace(replace(t.enrolled_period,'''','"'),'(','['),')',']') as js
  ) v
) ep;

编号 |注册期间 |开始日期 |结束日期 -: | :---------------------------- | :--------- | :--------- 1 | ('2021-10-11', '2021-10-31') | 2021-10-11 | 2021-10-31 编号 |注册期间 |期间 |开始日期 |结束日期 -: | :---------------------------- | :---------------------------- | :--------- | :--------- 1 | ('2021-10-11', '2021-10-31') | [“2021 年 10 月 11 日”，“2021 年 10 月 31 日”] | 2021-10-11 | 2021-10-31

db小提琴here

Answer 3

SELECT d.person_Id,
       d.University_Name,
       CAST(LEFT(d.Enrolled_period, CHARINDEX (',', d.Enrolled_period) - 1) AS DATE) AS Start_Date,
       CAST(RIGHT(d.Enrolled_period, CHARINDEX (',', d.Enrolled_period) - 1) AS DATE) AS End_Date
FROM
       (
           SELECT d.person_Id,
                  d.University_Name,
                  (REPLACE (REPLACE (REPLACE (d.Enrolled_period, '(', ''), ')', ''), '''', '')) AS Enrolled_period
           FROM
                  (
                      SELECT 24569 AS person_Id,
                             'Gothenburg university' AS University_Name,
                             '(''2007-08-09'',''2009-08-25'')' AS Enrolled_period
                      UNION
                      SELECT 24568 AS person_Id,
                             'Gothenburg university' AS University_Name,
                             '(''2019-07-09'',''2021-06-25'')' AS Enrolled_period
                  ) AS d
       ) AS d

Answer 4

你需要去掉“-”

select D.[start_date]
    ,D.[end_date]
from dbo.test_period T
    outer apply
    (
        select 
            replace(replace(replace(replace(replace(replace(
                T.enrolled_period
                ,'(', '')
                ,')', '')
                ,'''', '')
                ,',', '')
                ,'-', '')
                ,' ', '') as clean_period
    ) C
    outer apply
    (
        select
            cast(left(C.clean_period, 8) as date) as [start_date]
            ,cast(right(C.clean_period, 8) as date) as [end_date]
    ) D